Professional Documents
Culture Documents
Abstract
Background and methods: This paper addresses, in simulation, energy minimization of an autonomous electric
minibus operating in an urban environment. Two different case studies have been considered, each involving a
total of 10 different 2?km bus routes and two different average speeds. In the proposed method, the minibus
follows an optimized speed profile, generated using a genetic algorithm.
Results: In the first case study the vehicle was able to reduce its energy consumption by around 7 to 12% relative
to a baseline case in which it maintains a constant speed between stops, with short acceleration and deceleration
phases. In the second case study, involving mass variation (passengers entering and alighting) it was demonstrated
that the number of round trips that can be completed on a single battery charge is increased by around 10% using
the proposed method.
Keywords: Energy minimization, Autonomous vehicles, Electric minibuses, Speed profile optimization, Genetic
algorithms
© The Author(s). 2020 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made.
Torabi et al. European Transport Research Review (2020) 12:2 Page 2 of 8
model proposed in Genikomsakis and Mitrentsis [17]. This 3.2 Road profile
model, in addition to transforming the traction power re- In order to evaluate speed profiles in the simulation,
quirement at the wheels into battery power requirements a road profile is needed that represents the elevation
using a generic physics-based vehicle model, provides an as a function of longitudinal distance along the road.
estimation of the battery’s state-of-charge, while also con- The road profiles have been defined using composite
sidering the resistive forces such as aerodynamic drag and Bézier curves, i.e. a sequence of Bézier splines, as in
rolling resistance. Moreover, the model considers regenera- Torabi and Wahde [2]. Since only the longitudinal
tive braking system into the battery state of charge model. motion is considered, a road profile can be modelled
The energy consumption of the vehicle is calculated by in- using two-dimensional composite Bézier curves where
tegrating the following differential equation: the two dimensions are the longitudinal position and
elevation along the road. Thus, the road profiles are
dE=dt ¼ P tot ð1Þ represented in the following form:
where E is the energy consumption and Ptot is the total ðs; zÞ ≡ ðsi ðuÞ; zi ðuÞÞ ¼ P 0 ð1−uÞ3 þ 3P 1 uð1−uÞ2
power provided by the battery which is computed as þ3P 2 u2 ð1−uÞ þ P 3 u3 : i ¼ 0; …; n−1 (7)
Ptot ¼ Pb ηRTE ; ð2Þ where the vectors Pi are two-dimensional control points,
u is a parameter ranging from 0 to 1, and n is the total
Pb ¼ P m þ Paux : ð3Þ number of splines. With this representation, at any given
point s along the road, the corresponding spline index
Here, Pb denotes the battery’s output power. Pm repre- and the u-value can be determined. Therefore, the eleva-
sents the required power by the electric motor whereas tion can be written as z = z(s).
Paux is the auxiliary power, i.e. the power needed to op-
erate the vehicle’s auxiliary systems (a value that is con- 3.3 Speed profile
stant throughout the simulation). Moreover, ηRTE A speed profile is represented using two-dimensional
(slightly smaller than 1) represents the battery’s round composite Bézier curves, similar to the road profile rep-
trip efficiency, the precise value of which depends on resentation defined in Eq. (7) above, where the two di-
whether the battery is in charging or discharging mode; mensions are the longitudinal position and the
see Eq. (6) below. The battery’s state-of-charge (S) is up- longitudinal speed of the vehicle. An example is given in
dated as follows: Fig. 1, in which the stops can be clearly seen. The speed
profile provides a desired speed at any given position
dS=dt ¼ −P b =C; ð4Þ along the road, for the vehicle to follow using a PID con-
troller. Note that, when the vehicle is at a stop, and is
where C denotes the battery’s capacity (in Ws). Finally, about to leave, the desired speed is artificially inflated to
the electric motor’s required power is calculated by con- a small positive value, in order to start the motion. Once
sidering the longitudinal dynamics of the vehicle and its the vehicle’s speed is positive, it then follows the speed
engine efficiency: profile.
Pt ¼ v mð1 þ C i Þa þ F aero þ F g þ F roll ð5Þ 4 Methods
This section first describes the procedure for evaluating a
P t ηr ; if P t < 0 given speed profile. Next, the speed profile optimization
Pm ¼ ð6Þ
Pt =ηt ; if Pt > 0 method is described, and then the baseline method against
which the speed profile optimization results are compared.
where Pt is the traction power required from the electric
motor, v denotes the vehicle’s speed, m is the mass, Ci is 4.1 Evaluation of a speed profile
the correction factor for the rotational inertia force, and Assuming that a speed profile is available, the energy
a is the vehicle’s acceleration. The terms Faero, Fg and consumption of a vehicle can be measured by simulating
Froll represent the aerodynamic force, gravity force, and the motion of the vehicle as it follows the speed profile.
the rolling resistance force acting on the vehicle, respect- A simulation environment was written (in C#. NET)
ively. Finally, ηr and ηt are the electric motor’s efficiency implementing the vehicle model described in Sect. 3.1
in regenerative (charging) mode and in traction (dischar- and the road and speed profiles defined in Sects. 3.2 and
ging) mode, respectively. For a full description of the 3.3. To evaluate a speed profile performance in terms of
model and its parameter values, see Genikomsakis and energy consumption over a given road profile, the fol-
Mitrentsis [17]. lowing procedure is used: At each time step, the current
Torabi et al. European Transport Research Review (2020) 12:2 Page 4 of 8
Fig. 1 Top panel: A 2 km road profile with slope variations within ±1%. The boxes represent the bus stops along the route with the final stop at
the travelled distance of 2 km. The figure shows the complete profile for an entire drive cycle, from start to end, and back again. Bottom panel: A
corresponding optimized speed profile, for a case with an average speed between stops of 10 km/h; see Sect. 6.1
longitudinal location of the vehicle along the road is de- continuity (C1), which together ensure that the speed
termined, and it is then used to obtain the desired speed profile is continuous and smooth after decoding. In
(at the same location) from the speed profile. Thus, the order to meet these two requirements, the following
speed profile acts as a lookup table. The desired speed is conditions must hold:
then fed to the PID controller as its set speed. Note that
Piþ1;0 ¼ P i;3 f or i ¼ 0; …; n−1; ð8Þ
the set speed varies over a road profile. Therefore, due
to the controller’s non-zero response time, the vehicle Piþ1;1 −Piþ1;0 ¼ P i;3 −P i;2 f or i ¼ 0; …; n−1; ð9Þ
will not follow the set speed exactly, and neither is it re-
quired to do so in the method presented here. The pro- where Pi,j is the jth control point of the ith spline. More-
cedure is repeated, time step by time step, until the over, the speed at the start and the end of a profile
vehicle reaches the end of the road profile. should be set to 0. The first spline thus requires three
In this work, it is assumed that the motion of the ve- parameters, the following n-2 splines two parameters
hicle takes place in a dedicated lane so that it is able to each, and the final spline one parameter. Therefore, each
follow its speed profile. It is further assumed that there chromosome contains a total of 3 + 2(n-2) + 1 = 2n
are no traffic lights along the bus routes. However, if parameters.
traffic lights were present, they could be dealt with as For a given individual, the decoding procedure results
additional stops; see Sect. 7 for detailed discussion. in a speed profile that is evaluated as described in Sect.
4.1. The evaluation results in a single number, namely
4.2 Speed profile optimization the energy consumption Eend obtained when the vehicle
The speed profile optimization is carried out with respect reaches the end of the profile. If the maximum speed or
to energy consumption only, using a genetic algorithm the final battery state-of-charge constraints are violated,
(GA). A speed profile must fulfil certain constraints, the speed profile is assigned zero fitness. If not, the fit-
namely (i) the instantaneous maximum speed must not ness measure f is taken as
exceed an upper limit (the speed limit), (ii) the average 8 2
>
> v km
speed should not be below a certain user-defined thresh- < 1 e−cv v d −1 ; i f jv−vd j < 2
old, and (iii) the final battery state of charge (for the elec- f ¼ E end h ð10Þ
>
> km
tric vehicle) should be above a threshold to prevent : 0; i f jv−vd j > 2
h
battery damage.
As in any standard GA, the algorithm maintains a where v is the actual average speed, vd is the desired
population of M individuals (here 20), each defining a average speed (see Sect. 6), and cv is a positive constant.
speed profile. The speed profiles are encoded in chromo- Selection is carried out using standard tournament
somes that specify the speed values (i.e. the second di- selection with a given tournament size (typically 2–5)
mension of the control points in Eq. (7)), with the and a tournament probability of pt (around 0.7–0.8).
requirement of positional continuity (C0) and derivative Single-point crossover with a probability of pcross
Torabi et al. European Transport Research Review (2020) 12:2 Page 5 of 8
(around 0.7–0.9) is used. Once two new individuals considered. For each of the road profiles, the desired
are formed through selection and crossover, mutation average speed was set either to 10 km/h or 15 km/h,
is applied with probability of pmut (around 0.05–0.1) which are suitable values for cases where the distance
with either full-range mutation (with probability of between the bus stops are rather short; Oskarbski et al.
0.5) or creep mutation (with probability of 0.5) with a [19]. The maximum speed was set to 40 km/h. More-
creep rate of 10% of the full range. In order to make over, in all cases below, the vehicle’s battery initial state
sure that the best speed profile found so far is not of charge was set to 95% at the start of the trip and the
lost when forming the next generation (in the GA), mass of the vehicle (when empty) was set to 2000 kg. In
this profile is passed unchanged to the next gener- order to form a baseline case against which the
ation, a concept known as elitism. optimization algorithm can be compared, the energy
consumption of the bus was computed using the method
4.3 Baseline method described in Sect. 4.3 with same average speed, initial
The baseline case roughly mimics a typical driving se- battery state of charge, and mass. The performance is
quence performed by a bus driver, albeit executed in a typically measured either as the total energy saving (i.e.
more precise manner (with a PID controller), see e.g. E in Eq. (1); see also Dib et al. [12]), or the battery usage
Pourabdollah et al. [18]. In this case, a speed profile is savings (i.e. S in Eq. (4); see also Mensing et al. [9]).
defined with constant speed (referred to as the cruising Those two values differ slightly, due to the battery
speed) between stops. The first and last splines of a round-trip efficiency coefficient; see Eq. (2). Here, both
baseline speed profile define the acceleration and decel- measures will be given. In all cases (speed profile
eration phase, respectively. The cruising speed was set optimization and the baseline case), the speed profiles
so as to achieve a given average speed; see Sect. 6 below consisted of n = 40 splines so that each spline covered
for numerical values. An example of a baseline case is 50 m.
given in Fig. 2 below.
6.1 Case study I
5 Road data In this case study, the optimization was carried out for a
Ten synthetic road profiles of 2 km length were gener- single round trip, i.e. total of 4 km driving, for both of the
ated corresponding to short urban bus routes. The gen- bus route groups. For the purpose of comparison, the
erated bus routes were defined in such a way that their mass of the bus was fixed to the initial value (2000 kg)
slope variations were within ±1% for five routes (Route both for the baseline case and for the speed profile
group I) and within ±3% for the other five (Route group optimization.
II). There were four stops in total, at 0, 500, 1500, and The results obtained from the simulations (henceforth:
2000 m from the starting point; see also the top panel of Case study I) are presented in Table 1 for each of the
Fig. 1. The generated road profiles were similar to a bus bus route groups (Groups I and II). The second column
route in Kongsberg, Norway (considered within the from the left shows the required average speed between
framework of the Sohjoa Baltic project) which is roughly consecutive bus stops, and the two right-most columns
2 km long, with a total of 4 stops, and with small slope show the energy savings and the battery usage savings,
variation. The road profiles were defined with one spline respectively. Both the energy savings and the battery
for every 500 m, which is sufficient given the mild slope usage savings are calculated relative to the energy con-
variation. sumption and battery usage of the vehicle when follow-
ing the speed profiles defined using the baseline method.
6 Simulation results The results obtained from the simulations show that by
In order to evaluate the performance of the speed profile following the optimized speed profiles, the total energy
optimization method, two case studies have been consumption of the vehicle is reduced by an average of
Fig. 2 An example of a baseline speed profile for an entire driving cycle (from start to end, and back again). In this case, the average speed
between stops was set to 10 km/h
Torabi et al. European Transport Research Review (2020) 12:2 Page 6 of 8
Table 1 Simulation results obtained for Case study I. The second column from the left shows the average speed of the vehicle
between stops (set to either 10 or 15 km/h). The two right-most columns show the energy savings and battery usage savings
relative to the energy consumption and battery usage of the baseline case, respectively
Bus route Average speed (km/h) Energy savings (%) Battery usage savings (%)
Group I 10 12.62 ± 2.36 12.44 ± 2.69
15 10.19 ± 1.69 10.50 ± 1.57
Group II 10 9.50 ± 1.74 9.49 ± 1.95
15 7.40 ± 1.73 7.39 ± 1.95
around 11.5% and 8.5% for bus routes with slopes within m ¼ m0 þ n 75 kg ð11Þ
±1% (Group I) and ± 3% (Group II) respectively. More-
over, the battery usage of the vehicle is decreased by
similar values relative to the battery usage of the baseline where m0 is the initial mass (2000 kg) and n is the num-
method. ber of passengers chosen randomly from 0 to 6 (the
The bottom panel of Fig. 1 shows an example of an maximum number for the minibus considered here)
optimized speed profile for one of the routes in Group I, multiplied by the average weight of a passenger (here
in a case with 10 km/h average speed between stops. considered to be 75 kg). In order to make a fair compari-
Note that even though the speed profile gives an impres- son, the mass variation at stops was identical for both
sion of high speed variability, exaggerated by the com- methods (SPO and baseline). The time spent at the bus
pressed horizontal axis, it should be noted that the stops (which affects the total energy consumption) was
(magnitude of the) maximum acceleration is never above not included in the analysis, since it is the same for both
0.6 m/s2. Moreover, with 10 km/h average speed, the methods.
travel over the first 500 m (i.e. between the start and the The optimization was carried out for both bus
first stop) takes 3 min, implying that the 3–4 mild accel- routes (Groups I and II), stop-by-stop, generating
eration and deceleration cycles in that interval should one speed profile for each segment (i.e. the stretch
not cause discomfort to the passengers. Similar observa- of road between two stops) for every possible mass
tions can be made for the other intervals. value, and for each average speed value. As before,
the vehicle’s average speed was set either to 10 km/h
or 15 km/h between stops. At every stop during a
6.2 Case study II round trip, an optimized speed profile was selected
An additional case study (Case study II) was considered as by considering the change in the vehicle’s mass and
well in which the vehicle continues its operation until the the required average speed. The results obtained
battery state-of-charge drops below a certain threshold from the simulations are presented in Table 2. The
(here 20%). In this experiment, which is an extension of results show that by using the SPO method, the
Case study I, the vehicle’s motion was simulated using both electric vehicle is able to increase the number of
the speed profile optimization method and the baseline completed round trips by around 12% relative to the
method. In order to emulate a typical bus driving scenario, baseline method. Moreover, the energy and battery
the mass of the vehicle was varied at each bus stop in the usage savings when using SPO are similar to the
following way values obtained in Case study I.
Table 2 Simulation results obtained for Case study II. The second column from the left shows the average speed of the vehicle
between stops (set to either 10 or 15 km/h) whereas the third and the fourth columns give the average energy consumption of the
vehicle per round trip using the SPO method and the baseline method, respectively. The fifth and sixth columns from the left show
the number of completed round trips before the vehicle’s battery state-of-charge drops below 20%. The two right-most columns
show the average battery usage of the vehicle
Bus Average Average energy consumption per round trip Number of round trips before Average battery usage per round trip
route speed (kWh) discharge (%)
(km/h)
SPO baseline SPO baseline SPO baseline
Group I 10 1.013 ± 0.157 1.141 ± 0.186 37 ± 1 32 ± 1 2.03 ± 0.10 2.26 ± 0.03
15 1.015 ± 0.124 1.115 ± 0.167 42 ± 1 37 ± 1 2.01 ± 0.13 2.23 ± 0.19
Group II 10 1.074 ± 0.181 1.216 ± 0.193 36 ± 1 32 ± 1 2.08 ± 0.10 2.26 ± 0.04
15 1.120 ± 0.163 1.193 ± 0.183 43 ± 1 38 ± 1 2.18 ± 0.09 2.34 ± 0.03
Torabi et al. European Transport Research Review (2020) 12:2 Page 7 of 8
speed can be set to higher values as well, in order to fol- 4. Hellström, E., Åslund, J., & Nielsen, L. (2010). Design of an efficient algorithm
low a specific timetable. With a different average speed for fuel-optimal look-ahead control. Control Engineering Practice, 18(11),
1318–1327.
setting, one would then have to rerun the simulations to 5. Wirasingha, S. G., & Emadi, A. (2011). Classification and review of control
calculate the energy savings. strategies for plug-in hybrid electric vehicles. IEEE Transactions on Vehicular
A possible extension of this work would be to adapt Technology, 60(1), 111–122.
6. Moura, S. J., Fathy, H. K., Callaway, D. S., & Stein, J. L. (2011). A stochastic
the fitness measure (in the optimization) so that it in- optimal control approach for power management in plug-in hybrid electric
creases the emphasis on regenerative braking, thus vehicles. IEEE Transactions on Control Systems Technology, 19(3), 545–555.
resulting in larger energy savings, but possibly at the ex- 7. He, Y., Chowdhury, M., Pisu, P., & Ma, Y. (2012). An energy optimization
strategy for power-split drivetrain plug-in hybrid electric vehicles.
pense of slightly larger acceleration and deceleration. Transportation Research Part C: Emerging Technologies, 22, 29–41.
Another topic for future work would be to migrate the 8. Guo, L., Gao, B., Gao, Y., & Chen, H. (2017). Optimal energy management for
optimization procedure to a cloud-based server, as well HEVs in eco-driving applications using bi-level MPC. IEEE Transactions on
Intelligent Transportation Systems, 18(8), 2153–2162.
as running tests in real minibuses using the infrastruc- 9. Mensing, F., Trigui, R., & Bideaux, E. (2012). Vehicle trajectory optimization for
ture and vehicles available in the Sohjoa Baltic project. hybrid vehicles taking into account battery state-of-charge. In 2012 IEEE
vehicle power and propulsion conference (pp. 950–955).
10. Mets, K., Verschueren, T., Haerick, W., Develder, C., & De Turck, F. (2010).
8 Conclusion Optimizing smart energy control strategies for plug-in hybrid electric
For an autonomous electric bus with regenerative brak- vehicle charging. In 2010 IEEE/IFIP Network Operations and Management
Symposium Workshops (pp. 293–299).
ing operating in an urban environment, the proposed 11. Bashash, S., Moura, S. J., Forman, J. C., & Fathy, H. K. (2011). Plug-in hybrid
method provides energy savings of around 7 to 12% rela- electric vehicle charge pattern optimization for energy cost and battery
tive to a baseline case with a constant reference speed longevity. Journal of Power Sources, 196(1), 541–549.
12. Dib, W., Chasse, A., Mouline, P., Sciarretta, A., & Corde, G. (2014). Optimal
between stops. The savings translate to an increase of energy management for an electric vehicle in eco-driving applications.
around 10% in the number of round trips that can be ex- Control Engineering Practice, 29, 299–307.
ecuted on a single battery charge, corresponding roughly 13. Rohkämper, S., Hellwig, M., & Ritschel, W. (2017). Energy optimization for
electric vehicles using dynamic programming. In 2017 International
to an extra hour of operation for the case considered Conference on Research and Education in Mechatronics (REM) (pp. 1–5).
here. The results also demonstrate the applicability of 14. Jiang, X., Chen, L., Xu, X., Cai, Y., Li, Y., & Wang, W. (2018). Analysis and
the SPO method, originally developed for vehicles with optimization of energy efficiency for an electric vehicle with four
independent drive in-wheel motors. Advances in Mechanical Engineering,
combustion engines, to electric vehicles. 10(3) (pp. 1–9).
15. Maamria D., Gillet K., Colin G., Chamaillard Y. and Nouillant C., (2016). On the
Acknowledgements use of Dynamic Programming in eco-driving cycle computation for electric
The funding support of the EU Baltic Sea Interreg program through the vehicles. In 2016 IEEE Conference on Control Applications (CCA) (pp. 1288–
Sohjoa Baltic project (grant number R073) is gratefully acknowledged. 1293).
16. Rios-Torres, J., Sauras-Perez, P., Alfaro, R., Taiber, J., & Pisu, P. (2015). Eco-driving
Declarations system for energy efficient driving of an electric bus. SAE International Journal
The data used during the current study are available from the corresponding of Passenger Cars - Electronic and Electrical Systems, 8, 79–89.
author upon request. 17. Genikomsakis, K. N., & Mitrentsis, G. (2017). A computationally efficient
simulation model for estimating energy consumption of electric vehicles in
the context of route planning applications. Transportation Research Part D:
Authors’ contributions Transport and Environment, 50, 98–118.
All authors read and approved the final manuscript. 18. Pourabdollah, M., Bjärkvik, E., Fürer, F., Lindenberg, B., & Burgdorf, K. (2017).
Fuel economy assessment of semi-autonomous vehicles using measured
Funding data. In 2017 IEEE Transportation Electrification Conference and Expo (ITEC)
Open access funding provided by Chalmers University of Technology. (pp. 761–766).
19. Oskarbski, J., Birr, K., Miszewski, M., & Żarski, K. (2015). Estimating the average
Competing interests speed of public transport vehicles based on traffic control system data. In
The authors declare that they have no competing interests. 2015 IEEE International Conference on Models and Technologies for Intelligent
Transportation Systems (MT-ITS) (pp. 287–293).
Received: 3 July 2019 Accepted: 17 December 2019 20. Mersky, A. C., & Samaras, C. (2016). Fuel economy testing of autonomous
vehicles. Transportation Research Part C: Emerging Technologies, 65, 31–48.